Mysterious crash/freeze in X11 with newcons

[Matthew Rezny]

> I had two crashes, or more accurately, freezes, when web-browsing.
> 
> First was with Firefox on an easy National Weather Service page, I was
> scrolling down with down-arrow key, and system suddenly froze, no response
> to keyboard or mouse, display stayed (until I hit Reset).
> 
> uname -a shows
> 
> FreeBSD amelia4 11.0-CURRENT FreeBSD 11.0-CURRENT #4 r264310: Thu Apr 10
> 06:56:03 UTC 2014     root at amelia4:/usr/obj/usr/src/sys/SANDY11NC  amd64
> 
> NC in kernel name is for NewCons, SC would be SysCons, and I compiled kernel
> both ways.
> 
> Second was with Midori, actually twice, on a more complex site, 1040now.net
> :
> 
> I checked /var/log/messages, and /var/log/messages.0.bz2, and noticed
> 
> Apr 15 08:29:52 amelia4 kernel: error: [drm:pid806:drm_do_get_edid] *ERROR*
> VGA-1: EDID block 0 inva lid.
> 
> and a few minutes later,
> 
> Apr 15 08:33:58 amelia4 kernel: error: [drm:pid806:drm_do_get_edid] *ERROR*
> VGA-1: EDID block 0 invalid. Apr 15 08:33:58 amelia4 kernel: sysmouse:
> unknown ioctl: t:40007413
> Apr 15 08:33:58 amelia4 kernel: sysmouse: unknown ioctl: t:40007413
> Apr 15 08:33:58 amelia4 kernel: sysmouse: unknown ioctl: t:80007410
> 
> I notice from earlier, and think this must have been the crash with Firefox:
> 
> pr 14 23:09:18 amelia4 kernel: error: [drm:pid87853:drm_do_get_edid] *ERROR*
> VGA-1: EDID block 0 in valid.
> Apr 14 23:09:18 amelia4 kernel: sysmouse: unknown ioctl: t:40007413
> Apr 14 23:09:18 amelia4 kernel: sysmouse: unknown ioctl: t:40007413
> Apr 14 23:09:18 amelia4 kernel: sysmouse: unknown ioctl: t:80007410
> 
> 
> I think these errors must have been the cause of the freeze/crash, and
> looking for somebody who recognizes what is *ERROR* VGA-1: EDID block 0
> invalid.
> 
Due to proximity of the unknown ioctl, I assume that message is early in boot 
and not when the problem occurs. Invalid EDID means there was a problem 
getting the data from your monitor which tells the system what its 
capabilities are (resolutions and refresh rates). Either your monitor has 
invalid data, or you have a bad cable. Result is detection of available modes 
and choice of optimal mode cannot be done automatically. Your freezes are 
completely unrelated.

> Would I be better with syscons instead of newcons?  But then, how would I
> open a root window for such things as mounting and umounting
> disks/partitions, and anything else requiring root privilege?
> 
Open a root terminal (xterm, or whatever is your terminal of choice), or use 
su or sudo in an existing terminal. I don't think you will see any difference 
with syscons as this is an issue in X and not related to the console. If you 
beleived it might, it sholud be easy for you to test since you said you are 
building kernels with both. If you use vesa driver with syscons and set a 
graphical mode, you can get back to console (useful if X has frozen but the 
kernel is alive), but you'll never be able to return to X again (even under 
normal circumstances).

> Or would I do better on 10-stable?  But then if the problem is with the new
> Xorg, 10-stable would be no stabler than the bleeding-edge 11-head.
> 
> I am/was in window manager icewm-1.3.8 but think, since the problem occurred
> with both Firefox and Midori, the problem is with the new Xorg and possibly
> newcons rather than web browser.
> 
> Reason for building www/midori was to see if it would be safer than Firefox
> in terms of stability, and to compare bugs with Midori-0.5.5 on the System
> Rescue CD.
> 
> Tom

I had many freezes when first using new Xorg with KMS drivers. Firefox was by 
far the worst trigger, but not the source of the problem. In general, any GTK 
apps could trigger the hard freezes more easily than Qt apps. Upgrading cairo 
was my first try, but little changed. Upgrading Mesa to 10.1 made a world of 
difference. I still see pauses (all of X hangs for a matter of seconds) 
triggered by Firefox, but the total freeze and catastrophic screen corruptions 
are essentially eliminated.

You have not said what video hardware you are using. If you are using Radeon, 
then definitely try Mesa 10.1 from the xorg-dev repo experimental branch. If 
you are using Intel, you are stuck with Mesa 9.1.7 until the kernel drm bits 
are updated.